In the current reduced network formulation of the transient energy function method, all system loads are conventionally modelled as constant-impedance load to reduce the network to the internal nodes of the generators. Neglecting other types of load model (e.g. constant-power and constant-current load models), as in current approaches, may lead to incorrect results. The paper describes the inclusion of load modelling in the transient energy function method, using a sparse network formulation. In this formulation, the loads can be represented as they occur in practice (constant-impedance, constant-power, constant-current, or any combination thereof). Applications of the new technique to practical utility systems have proved the effectiveness of the new technique in producing more accurate results. The results compare favourably with those obtained by the time domain solution techniques.